A short-term and short-time integrated hail weather warning method for a FAST area

Abstract

The application discloses a short-term and short-time integrated hail weather warning method for a FAST area, relates to the technical field of hail weather warning, and comprises the following steps: selecting sounding data of the nearest sounding time before hail occurs to perform warning index statistics, and obtaining a statistical result; constructing a warning index threshold value according to the statistical result; screening an initial sample set of a hail weather circulation situation feature according to the warning index threshold value, and obtaining a screened sample set; constructing an initial short-term prediction model of hail weather; testing an optimized short-term prediction model according to historical hail weather radar data and the screened sample set, and outputting a best hail weather prediction quantity threshold value; and issuing a hail short-time warning and a hail short-term warning according to the best hail weather and preset strong convective cell identification conditions. The application significantly improves the hail warning capability in a complex terrain area.

Description

Technical Field

[0001] This invention relates to the field of hail weather warning technology, and in particular to a short-term, short-duration integrated hail weather warning method for the FAST area. Background Technology

[0002] Currently available hail forecasting methods for the FAST region largely rely on satellite and radar data, resulting in short forecast lead times (≤1 hour), which is insufficient to meet the advance deployment requirements of hail prevention operations. Although some studies have attempted to introduce radiosonde data to establish short-term forecast indicators, a systematic and operationally integrated forecasting model has not yet been formed. In particular, there is a lack of in-depth analysis of the triggering mechanisms of hail processes in the unique terrain of the FAST region, and a failure to establish differentiated warning thresholds based on weather situation classifications, leading to limitations in forecast accuracy and timeliness.

[0003] Therefore, a short-term, short-duration integrated hail weather warning method for the FAST area was developed to solve the above problems. Summary of the Invention

[0004] This invention proposes a short-term and short-duration integrated hail weather warning method for the FAST region to solve the problem of low forecast accuracy and timeliness caused by the inability of existing technologies to achieve integrated short-term and short-duration warnings.

[0005] The present invention achieves the above objectives through the following technical solutions:

[0006] This invention provides a hail weather prediction method for the FAST region, comprising:

[0007] Data was acquired, including real-time weather radar data for the target area, as well as conventional ground observation data, radiosonde data, and ERA5 reanalysis data for the same historical time period.

[0008] The radiosonde data from the most recent radiosonde time before the hail occurred were selected to statistically analyze the early warning indicators and obtain the statistical results.

[0009] Early warning indicator thresholds were constructed based on statistical results;

[0010] The initial sample set of hail weather circulation characteristics in conventional ground observation data is filtered according to the warning index threshold to obtain the filtered sample set;

[0011] An initial short-term forecast model for hail weather was constructed. The initial short-term forecast model is a linear regression equation. The dependent variable of the linear regression equation is the forecast amount of hail weather, and the independent variable is the warning indicator.

[0012] The significance of the initial short-term forecast model was tested using the stepwise regression method, and the optimized short-term forecast model was obtained.

[0013] The short-term forecast model is tested based on historical hail weather radar data from conventional ground observation data and a selected sample set, and the optimal hail weather forecast threshold is output.

[0014] When the real-time output exceeds the forecast threshold for optimal hail weather, a short-term hail warning is issued. Then, it is further determined whether the real-time weather radar data meets the preset strong convective cell identification conditions. If so, a short-term hail warning is issued.

[0015] Furthermore, the method for determining the early warning indicators includes:

[0016] Quality control of historical hail weather radar data;

[0017] Based on conventional ground observation data, ERA5 reanalysis data, and quality-controlled hail weather radar data, historical hail weather processes were observed and analyzed, simulated, and tested for terrain sensitivity to obtain the analysis results.

[0018] Based on the analysis results and meteorological principles, ten physical quantities were selected as early warning indicators for hail weather in the FAST area, taking into account thermal instability, water vapor conditions, energy conditions, dynamic conditions, and special layer height.

[0019] The ten physical quantities include the Sackler index, K index, lifting index, storm intensity index, convective available potential energy, 0°C layer height, 0~-20°C layer thickness, 0~6km deep vertical wind shear, 500~700hPa temperature difference, and 700hPa specific humidity.

[0020] Furthermore, based on the statistical results, early warning indicator thresholds are constructed, including:

[0021] Historical hail weather radar data are classified to obtain cold air intrusion processes, favorable convergence patterns, and post-frontal elevated thunderstorm processes.

[0022] Based on the classification results, outliers caused by post-frontal elevated thunderstorm processes were excluded from the statistical results.

[0023] Then, based on the statistical results after excluding outliers, a threshold for early warning indicators is constructed.

[0024] Furthermore, the warning indicator thresholds include: Saud index < 0.05, K index > 24, lifting index < 0, storm intensity index > 260, convective effective potential energy > 360 J / kg, 0℃ layer height > 3 km and < 5 km, 0~-20℃ layer thickness > 2.6 km, 0~6 km vertical wind shear > 15 m / s, 500~700 hPa temperature difference > 15℃, and 700 hPa specific humidity > 5 g / kg.

[0025] Furthermore, an initial sample set of hail weather circulation patterns from routine ground observation data was screened based on early warning indicator thresholds, including:

[0026] In the initial sample set of hail weather circulation patterns (e.g., routine ground observation data of hail weather in a certain period of a certain year in the FAST area), samples that meet the following conditions are initially screened out: there is obvious northerly cold air intrusion into the target area on weather maps at 500 hPa or below; the target area is controlled by a low-pressure center at 850 hPa when an upper-level trough exists; there is a Kunming quasi-stationary front; and radiosonde data shows the presence of a deep inversion layer.

[0027] The initially selected samples are substituted into the ten early warning indicators for testing. When the number of physical quantities that meet at least the threshold of the early warning indicator is 7, it indicates that the corresponding sample has passed the test, and the selected sample set is obtained.

[0028] Furthermore, based on the selected sample set, ten early warning indicators are used as independent variables, and the forecast amount of hail weather is used as the dependent variable. The selected sample set, independent variables, and dependent variables are input into the multiple linear regression module, and the least squares method is used for fitting. The constant term C, the regression coefficients A1 to A10 of the ten early warning indicators, and the t-statistics and p-values ​​of each coefficient are output to obtain the initial short-term forecast model, the expression of which is:

[0029] ;

[0030] Where Y is the dependent variable, i.e., the forecast amount of hail weather, and C is a constant. The regression coefficients of the ten early warning indicators are given. The independent variables are ten early warning indicators, namely, the Saud index, the K index, the lifting index, the storm intensity index, the convective effective potential energy, the 0°C layer height, the 0~-20°C layer thickness, the 0~6km deep vertical wind shear, the 500~700hPa temperature difference, and the 700hPa specific humidity. The regression coefficients of the ten early warning indicators are all in the range of [-1, 1].

[0031] A1-A 10 The constant C is not a fixed value, but is the result automatically calculated after "training" or "fitting" based on the "screened sample set" unique to this invention and through the standard linear regression calculation process (i.e., least squares method).

[0032] Furthermore, the initial short-term forecast model is subjected to a significance test based on the stepwise regression method to obtain the optimized short-term forecast model, including:

[0033] Based on the above fitting results, the stepwise regression module in the statsmodels library is called. The significance is determined by the t-statistic and p-value of each independent variable in the initial short-term forecast model. The significance level for the introduced variables is set to p < 0.05, and the removal level is p > 0.01. After iterative execution of the algorithm, the optimized model is automatically output. The optimized short-term forecast model is as follows:

[0034] ;

[0035] Where X1, X2, X3, and X4 represent the vertical wind shear of 0–6 km, the lifting index, the temperature difference of 500–700 hPa, and the storm intensity index, respectively; and A1, A2, A3, and A4 represent the regression coefficients of the vertical wind shear of 0–6 km, the lifting index, the temperature difference of 500–700 hPa, and the storm intensity index, respectively.

[0036] Statsmodels is a Python library focused on statistical modeling and inference, providing tools for linear regression, time series analysis, hypothesis testing, and more.

[0037] Furthermore, the short-term forecast model is tested based on historical hail weather radar data from conventional ground observation data and a selected sample set, and the optimal hail weather forecast threshold is output, including:

[0038] The historical hail weather radar data after quality control and the selected sample set were substituted into the short-term forecast model;

[0039] A formula for calculating the discrete true value of hail weather is constructed based on binary classification labels. The formula for calculating the discrete true value is as follows:

[0040] ;

[0041] Y i These are the discrete true values ​​for hail weather.

[0042] Compare the continuous forecast value Y calculated by the short-term forecast model with the discrete true value Y. i The accuracy rate of each preset forecast threshold to be determined is statistically analyzed to obtain the optimal forecast threshold for hail weather.

[0043] Using established hail weather forecasting models The continuous forecast for each sample is calculated by performing calculations on all historical samples used for modeling and validation. ;

[0044] Define a threshold search interval (from the minimum to the maximum value of Y) that covers a reasonable range of the predicted value Y. Iterate through each candidate threshold T within this interval with a step size of 0.05. For each T, perform the following operations:

[0045] (1) Classification: All Y i The sample of >T was judged as "hail forecast", and Y was... i Samples with a value ≤ T are classified as "no hail forecast".

[0046] (2) Calculate performance indicators: Based on the classification results and the actual category of the sample (hail / non-hail), calculate key business performance indicators, which mainly include: accuracy (the proportion of samples that are correctly predicted), false negative rate (the proportion of hail that actually occurs but is predicted as no hail), and false alarm rate (the proportion of no hail that is predicted as hail).

[0047] From all candidate thresholds, select a subset of thresholds that meet the condition of "false negative rate = 0", and then select the threshold with the highest accuracy from this subset as the final optimal discrimination threshold.

[0048] Furthermore, quality control includes:

[0049] Differential phase shift de-folding: This involves checking the continuity of the radial differential propagation phase shift before and after historical hail weather radar data to ensure that the measurable range is limited to a 180° interval.

[0050] Filtering: This involves performing comprehensive wavelet denoising on the defolded differential propagation phase shift and calculating the differential propagation phase shift rate using the least squares fitting method.

[0051] Attenuation correction: An adaptive constraint algorithm is used to correct the attenuation of the horizontal reflectivity factor and the differential reflectivity factor, and adaptive Barnes interpolation is performed on various radar parameters.

[0052] Furthermore, the preset conditions for identifying strong convective cells include: combined reflectivity ≥ 45 dBZ; vertical cumulative liquid water content > 10 kg / m²; particle classification identifying high-density graupel or hail; and obvious mesocyclone characteristics, i.e., shear velocity ≥ 15 m / s.

[0053] The beneficial effects of this invention are as follows:

[0054] This invention presents a short-term and short-duration integrated hail weather warning method for the FAST region. By combining weather pattern classification, physical quantity diagnosis, multivariate stepwise regression, and real-time radar identification, it achieves seamless forecasting from short-term (12–24 hours) to short-duration (0–2 hours). This invention particularly emphasizes terrain sensitivity analysis and classification threshold setting, significantly improving hail warning capabilities in complex terrain areas. Attached Figure Description

[0055] Figure 1 This is a technical roadmap of the short-term and short-duration integrated forecast model for hail weather in the FAST area according to an embodiment of the present invention;

[0056] Figure 2 These are the combined reflectance maps of 18:02, 18:46, and 19:15 in the embodiments of the present invention, respectively. Figure 2 (a), (b), and (c) in the text;

[0057] Figure 3 These are the reflectance factors of 18:02, 18:46, and 19:15 in the embodiments of the present invention, respectively. Figure 3 (d), (e), and (f) in the text;

[0058] Figure 4 These are particle recognition cross-sectional views at 18:02, 18:46, and 19:15 in the embodiments of the present invention, respectively. Figure 4 (g), (h), and (i) in the text;

[0059] Figure 5 This is a combined reflectance profile of the mode simulation at 10:30 UTC in an embodiment of the present invention;

[0060] Figure 6 This is a reflectance factor profile of the mode simulation at 10:30 UTC in an embodiment of the present invention;

[0061] Figure 7 This is a combined reflectance profile of the mode simulation at 11:30 UTC in an embodiment of the present invention;

[0062] Figure 8 This is a reflectance factor profile of the mode simulation at 11:30 UTC in an embodiment of the present invention;

[0063] Figure 9 This is a combined reflectance profile of the mode simulation at 12:15 UTC in an embodiment of the present invention;

[0064] Figure 10 This is a reflectance factor profile of the 12:15 UTC mode simulation in this embodiment of the invention. Detailed Implementation

[0065] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0066] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0067] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0068] like Figure 1 As shown, the technical approach for short-term and short-duration hail weather forecasting models based on the multivariate stepwise regression method includes the following steps:

[0069] S101, collect relevant data, including:

[0070] (1) Conventional ground observation data, including historical hail weather radar data and an initial sample set of hail weather circulation patterns;

[0071] (2) Radiosonde data of the target area at 8:00 and 20:00;

[0072] (3) ERA5 reanalysis data of the target area;

[0073] (4) Real-time weather radar data of the target area, taking X-band dual-polarization radar data as an example, such as reflectivity (ZH), differential propagation phase shift (ΦDP), differential reflectivity (ZDR), differential propagation phase shift rate (KDP), zero hysteresis cross-correlation coefficient (ρHV) within the detection range. The parameters of X-band dual-polarization radar data are shown in Table 1:

[0074] Table 1

[0075]

[0076] S102 performs quality control on individual radar data collected by S101 based on hail weather cases recorded from conventional ground observation data. This includes defolding, filtering, and attenuation correction.

[0077] S103, a typical hail weather event that occurred in the FAST area was selected, and observational analysis, simulation analysis (WRF-LES model) and terrain sensitivity test were carried out on it.

[0078] S104, after analyzing the weather patterns of the 28 hail events collected in S102 in the FAST region, is divided into three categories:

[0079] Cold air intrusion type: Significant cold air masses in the middle and upper levels move southward along different paths, intruding into the warm and moist air layer over Guizhou, thus triggering strong convection in favorable areas. This type is primarily driven by deep cold high-pressure systems or the airflow behind cold vortices and cyclones; strong cold advection is the main lifting triggering mechanism. This type of hail often occurs at surface convergence lines, dew point fronts, or in areas ahead of moving cold fronts, and is a typical example of surface-triggered convection.

[0080] Favorable convergence type: This type of hail weather is triggered by sustained and strong dynamic convergence and uplift caused mainly by weather systems such as low-level vortices and shear lines, against a backdrop of weak cold air activity. Low-level convergence is the core triggering mechanism, and the location and orientation of the hail landing area are consistent with the surface convergence line, the 700hPa low-level jet stream axis, or the low-level shear line.

[0081] Post-frontal elevated thunderstorm type: This is a special type of hail weather that occurs above a near-surface cold air cushion and is triggered by mid-level warm and moist advection and lifting mechanisms. Its convection occurs above the inversion layer and depends on the near-surface cold cushion formed by the Kunming quasi-stationary front or a southward-moving cold front; the convection is triggered by mid-level dynamic lifting, and the hail area is aligned with the direction of the 700hPa jet stream axis or the 700hPa temperature ridge.

[0082] This step aims to refine the causal mechanisms of hail weather, providing differentiated processing strategies for subsequent steps. Particularly in S107, classification directly leads to differentiated warning threshold settings, enhancing the model's adaptability and accuracy. This classification method enables the integrated model to handle both common and special types (such as post-frontal elevated thunderstorms), improving the timeliness and effectiveness of disaster prevention and mitigation.

[0083] Based on meteorological principles and the analysis results of S103, ten physical quantities were selected as early warning indicators for hail weather in the FAST area, taking into account thermal instability, water vapor conditions, energy conditions, dynamic conditions, and special layer heights. These are: Sark Index (SI), K Index, Lift Index (LI), Storm Intensity Index (SSI), Convective Available Potential Energy (CAPE), 0℃ layer height, 0~-20℃ layer thickness, 0~6km deep vertical wind shear, 500~700hPa temperature difference, and 700hPa specific humidity.

[0084] S106: The most recent radiosonde time before the hailstorm was selected for the hailstorm process. Using the radiosonde data from Guiyang station collected in S101, ten physical quantities selected in S105 were statistically analyzed, including the vertical wind shear at a depth of 0-6 km. Calculate using equation (1):

[0085] (1);

[0086] In the above formula, V1 is the wind speed near the ground, V2 is the wind speed at a height of 6km, and D is the wind direction difference from the ground to a height of 0-6km.

[0087] S107. Based on the statistical results of each physical quantity in S106, and after excluding the outliers caused by the post-frontal elevated thunderstorm process, the physical quantity warning index thresholds for the FAST area are established as follows: SI index < 0.05, K index > 24, LI index < 0, SSI index > 260, CAPE > 360 J / kg, 3km < 0℃ layer height < 5km, 0~-20℃ layer thickness > 2.6km, 0~6km vertical wind shear > 15m / s, 500~700hPa temperature difference > 15℃, 700hPa specific humidity > 5g / kg;

[0088] Post-frontal elevated thunderstorms typically occur after a cold front has passed, under conditions where a quasi-stationary front forms in Kunming. This type of hail event is difficult to forecast. Its TlogP chart shows extremely low convective effective potential energy and a high Schaeffler index. The weather situation appears stable, but there is a thick inversion layer and an unstable layer in the middle and lower troposphere. This results in the SI index often showing extremely high positive values, the CAPE often being extremely low or even 0, and the K and LI indices also showing similar "pseudo-normal" values.

[0089] S108 identified 23 samples ("test set") from routine surface observation data from January to April 2022 that met the criteria for hail weather circulation patterns in the FAST region. The test set assessed whether each physical quantity in these samples met the ten warning indicator thresholds established in S107, thus validating the model's effectiveness. The core test was determining "how many indicators a sample simultaneously meets for the best warning effect." The results showed that the best warning effect was achieved when at least seven physical quantities met the required warning indicator thresholds. These samples may contain different types; the classification results from S104 helped ensure the representativeness of the samples and considered the differences in warning thresholds across different types during the testing process, thereby optimizing the accuracy of the warning model. Table 2 shows the warning indicator tests for individual cases meeting the hail weather circulation patterns in the FAST region from January to April 2022.

[0090] Table 2

[0091]

[0092] The specific requirements for hail weather circulation patterns are as follows: there is a clear northerly cold air intrusion into Guizhou region on weather maps at or below 500 hPa; Guizhou region is controlled by a low-pressure center at 850 hPa when an upper-level trough exists; there is a Kunming quasi-stationary front; and radiosonde data in the target area show the presence of a deep inversion layer.

[0093] S109, set the dependent variable, that is, perform binary quantitative processing on the occurrence of hail weather process, construct the discrete true value calculation formula of hail weather based on binary classification label, and the mathematical expression is shown in (2); construct the short-term forecast model of hail weather as shown in (3).

[0094] (2);

[0095] In this invention, the prediction objective is "whether hail occurred." The formula means that, in the historical dataset used for model training and testing, for each weather process sample i, if hail was indeed verified to have occurred, it is labeled Y. i The value is assigned to 1 if hail does not occur; otherwise, the value is assigned to 0. In S1012, the verification of the model's forecast effectiveness is based on this definition.

[0096] (3);

[0097] In the above formula, Y is the dependent variable, i.e., the forecast quantity, and C is a constant. The regression coefficients for the ten physical quantity indicators are: The independent variables are ten physical quantities. The work of S108 ensured that these ten physical quantities were preliminary screening candidates that were highly correlated with hail occurrence.

[0098] S1010, calculate the t-statistic and p-value of each independent variable in the initial model in the above steps to determine significance. Set p>0.01 as the standard value for removing the model. When p<0.05, the independent variable passes the significance test. Use the F-test method to test the significance of the final model as a whole, and calculate the p-value in the same way. The results show that the calculated model p=0.000<0.05, which proves that the model has good performance.

[0099] The short-term forecast equation for hail weather established by the multivariate stepwise regression method of S1010 is shown in equation (4):

[0100] (4);

[0101] In the above formula, X1, X2, X3, and X4 represent the vertical wind shear of 0–6 km, the LI index, the temperature difference of 500–700 hPa, and the SSI index, respectively. The coefficients of the four indicators show that the vertical wind shear of 0–6 km and the LI index have a significant negative impact on the forecast amount Y, while the temperature difference of 500–700 hPa and the SSI index have a significant negative impact on the forecast amount Y.

[0102] S1012, the 28 hail weather cases in the FAST region selected by S102 and the 23 samples from January to April 2022 selected by S108 that meet the characteristics of hail weather circulation are substituted into the short-term forecast model output by S1011 for testing. By comparing the continuous forecast value Y calculated by the short-term forecast model with the discrete true value Yi determined by formula (2), the accuracy of the model can be objectively evaluated.

[0103] The forecast values ​​were divided into three ranges: >0.2, >0.3, and >0.4. The results showed that using >0.3 as the forecast value range threshold for the short-term hail forecast model yielded the best forecast performance. Step S108 determined the optimal quantitative thresholds for the qualitative indicators (7 items), while this step determined the optimal numerical threshold (0.3) for the quantitative model output. Both steps optimized the model's decision critical points through historical data testing to ensure the best operational application results.

[0104] S1013 When the short-term forecast model outputs Y>0.3, based on real-time data from the X-band dual-polarization radar, it identifies strong convective cells that meet any of the conditions, issues short-term hail warnings for 0–2 hours, and tracks their movement path and intensity changes in real time.

[0105] In one embodiment, S102 specifically includes the following steps:

[0106] S21, Differential Phase Shift Defolding: This involves checking the continuity of the radial differential propagation phase shift φDP before and after the transition to ensure that the measurable range is limited to a 180° interval.

[0107] S22. Filtering: This involves performing comprehensive wavelet denoising on the defolded differential propagation phase shift φDP and calculating the differential propagation phase shift rate (KDP) using the least squares fitting method.

[0108] S23. Attenuation Correction: The horizontal reflectivity factor (ZH) and differential reflectivity factor (ZDR) are corrected for attenuation using the adaptive constraint algorithm (SCWC), and adaptive Barnes interpolation is performed on various radar parameters.

[0109] In one embodiment, S103 specifically includes the following steps:

[0110] S31. Conduct observation and analysis, including circulation pattern characteristics analysis, thermodynamic and water vapor characteristics analysis, and radar echo evolution characteristics analysis.

[0111] The simulation parameter settings for the three-layer nested design of S32 and WRF-LES modes are shown in Table 3:

[0112] Table 3

[0113]

[0114] S33. The terrain sensitivity test only changes the terrain height of the area to drive the WRF-LES mode. In terms of terrain settings, the terrain height in the simulation area is uniformly set to 500m.

[0115] In one embodiment, S1010 specifically includes the following steps:

[0116] S101. Regress the ten physical quantity indicators with the predicted quantity, with the predicted quantity as the dependent variable and the ten physical quantity indicators as the independent variables. The item with the largest p value is the physical quantity of the -20℃ layer, p=0.841, which is much greater than the threshold requirement of 0.05. Therefore, it is removed from the model.

[0117] S102. Regress the remaining 9 independent and dependent variables again, and continue to eliminate insignificant independent variables;

[0118] S103, finally retain four independent variables: vertical wind shear from 0 to 6 km, LI index, temperature difference from 500 to 700 hPa, and SSI index.

[0119] In one embodiment, S1013 specifically includes the following steps:

[0120] The criteria for identifying strong convective cells include:

[0121] S131, Combined reflectance ≥45dBZ;

[0122] S132, Vertical cumulative liquid water content (VIL) > 10 kg / m²;

[0123] S133, Particle classification identifies high-density graupel (HDG) or hail (RH);

[0124] S134 exhibits distinct mesocyclone characteristics (shear velocity ≥ 15 m / s).

[0125] To demonstrate the inventiveness and technical value of the present invention, this section describes the application of the technical solution in specific products or related technologies.

[0126] After quality control of the radar data of FAST area on January 4, 2022, it was determined whether the embodiment was a hail weather process. The hail formation mechanism and topographic effect were analyzed by observation and simulation. Then, the four indicators were substituted into the hail weather forecast equation established by the multivariate stepwise regression method, and the forecast quantity Y was calculated to determine whether the embodiment should carry out hail prevention operations immediately.

[0127] Evidence of the effects of the embodiments. The embodiments of the present invention have achieved a series of significant effects during research and development and application, demonstrating clear advantages over existing technologies. The following, combined with data and charts from the experimental process, elaborates in detail on the specific manifestations of these superiorities.

[0128] The experimental supercell hailstorm occurred on January 4, 2022, under favorable circulation conditions of an eastward-moving plateau trough and southward-moving cold air, classifying it as a cold air intrusion-type hail event. The hail cells had relatively long lifespans, with a maximum hail diameter of 20 mm and a maximum hail density reaching 400 hailstones / m³. 2 .

[0129] Typically, cold air intrusion-type hailstorms are characterized by three intrusion paths: northwest, north, and northeast. The most common weather systems causing cold air intrusions are deep high-pressure systems, including the Siberian-Mongolian High, blocking highs, and the Balkshy High. Other systems include Mongolian cyclones and the Siberian cold vortex. In Guizhou, the region is usually controlled by a thermal low. Systems that play an important role in favorable convergence-type hailstorms include thermal lows, southwest vortices, plateau vortices, and low-level shear lines. Post-frontal elevated thunderstorms usually occur after a cold front passes, under conditions where a quasi-stationary front forms in Kunming. This type of hailstorm is difficult to forecast. Its T-logP chart shows very low convective effective potential energy and a high Schaeffler index. The weather pattern appears stable, but there is a thick inversion layer and an unstable layer in the middle and lower levels.

[0130] like Figure 2-4 The figure shows the combined reflectivity, reflectivity factor, and particle identification profiles at 18:02, 18:46, and 19:15 during the experiment. At 18:02, after being lifted by the terrain, the developing echo cell condenses a large amount of water vapor, forming a supercooled water accumulation zone at an altitude of 3-4 km. Ice crystals adhere to supercooled raindrops, forming a large number of LDG (low-density graupel) particles, providing hail embryos for hail formation. At 18:46, LDG adheres to supercooled water, forming a large amount of RH (reflectance), at which point the combined reflectivity of the echo cell reaches its strongest, resulting in hail. At 19:15, the echo cell is in the dissipation stage. After the hail settles, the echo intensity decreases significantly, but because the echo cell moves uphill at this time, the cloud body strengthens, which is conducive to the formation of ice phase particles. Therefore, a large number of LDG and HDG (high-density graupel) particles still exist in the cloud.

[0131] like Figure 5-10The figure shows the combined reflectance and reflectance factor profiles of the model simulation at 10:30, 11:30, and 12:15. Compared with the observation results, the reduction of the single hail cell is not significant because the simulation results show the characteristics of a multi-cell hailstorm. The newly generated strong convective cell interacts with the original cell, affecting the characteristics of its dissipation phase. The echo intensity, movement direction, vertical development height, and generation time and location of the hailstorm cell in the model results are similar to the observation results in the three stages. Therefore, the simulation effect is considered to be good. The main deviation is that the hailstorm cell exists independently in the observation results, while it shows the characteristics of a multi-cell hailstorm in the simulation results. The established hail weather forecast model was used for verification. The calculated single-cell forecast quantity Y>0.3 indicates that a short-term hail warning should be issued. Then, combined with a combined reflectance ≥45dBZ, a large amount of RH and HDG particles were identified, and a short-term hail warning should be issued.

[0132] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A short-term, short-duration integrated hail weather warning method for the FAST region, characterized in that, include: Data was acquired, including real-time weather radar data for the target area, as well as conventional ground observation data, radiosonde data, and ERA5 reanalysis data for the same historical time period. The radiosonde data from the most recent radiosonde time before the hail occurred were selected to statistically analyze the early warning indicators and obtain the statistical results. Early warning indicator thresholds were constructed based on statistical results; The initial sample set of hail weather circulation patterns in conventional ground observation data was filtered based on the warning index thresholds to obtain the filtered sample set; An initial short-term forecast model for hail weather was constructed. The initial short-term forecast model is a linear regression equation. The dependent variable of the linear regression equation is the forecast amount of hail weather, and the independent variable is the warning indicator. The significance of the initial short-term forecast model was tested using the stepwise regression method, and the optimized short-term forecast model was obtained. The short-term forecast model is tested based on historical hail weather radar data from conventional ground observation data and a selected sample set, and the optimal hail weather forecast threshold is output. Real-time weather radar data is input into the short-term forecast model to obtain real-time output. When the real-time output is greater than the forecast threshold for the best hail weather, a short-term hail warning is issued. Then, it is further determined whether the real-time weather radar data meets the preset strong convective cell identification conditions. If so, a short-term hail warning is issued. The method for determining the early warning indicators includes: Quality control of historical hail weather radar data; Based on conventional ground observation data, ERA5 reanalysis data, and quality-controlled hail weather radar data, we conducted observational analysis, simulation analysis, and terrain sensitivity testing on hail weather processes to obtain analysis results. Based on the analysis results and meteorological principles, ten physical quantities were selected as warning indicators for hail weather in the FAST area, taking into account thermal instability, water vapor conditions, energy conditions, dynamic conditions, and special layer heights. These indicators include the Schaeffler index, K index, lifting index, storm intensity index, convective available potential energy, 0℃ layer height, 0~-20℃ layer thickness, 0~6km deep vertical wind shear, 500~700hPa temperature difference, and 700hPa specific humidity. Early warning indicator thresholds are constructed based on statistical results, including: Historical hail weather radar data are classified to obtain cold air intrusion processes, favorable convergence patterns, and post-frontal elevated thunderstorm processes. Based on the classification results, outliers caused by post-frontal elevated thunderstorm processes were excluded from the statistical results. Then, based on the statistical results after excluding outliers, a threshold for early warning indicators is constructed.

2. The short-term, short-duration integrated hail weather warning method for the FAST area according to claim 1, characterized in that, The warning threshold indicators include: Saud index < 0.05, K index > 24, lifting index < 0, storm intensity index > 260, convective effective potential energy > 360 J / kg, 0℃ layer height > 3 km and < 5 km, 0~-20℃ layer thickness > 2.6 km, 0~6 km vertical wind shear > 15 m / s, 500~700 hPa temperature difference > 15℃, and 700 hPa specific humidity > 5 g / kg.

3. The short-term, short-duration integrated hail weather warning method for the FAST area according to claim 1, characterized in that, The initial sample set of hail weather circulation patterns in routine ground observation data was screened based on early warning index thresholds, including: Initially, samples meeting certain conditions were selected from the initial sample set of hail weather circulation patterns. These conditions included: a clear northerly cold air intrusion into the target area on weather maps at 500 hPa or below; the target area being controlled by a low-pressure center at 850 hPa when an upper-level trough was present; the presence of the Kunming quasi-stationary front; and radiosonde data showing the presence of a deep inversion layer. The initially selected samples are substituted into the ten early warning indicators for testing. When the number of physical quantities that meet at least the threshold of the early warning indicator is 7, it indicates that the corresponding sample has passed the test, and the selected sample set is obtained.

4. The short-term, short-duration integrated hail weather warning method for the FAST area according to claim 1, characterized in that, Constructing an initial short-term forecast model for hail weather, including: Based on the selected sample set, ten early warning indicators are used as independent variables, and the forecast amount of hail weather is used as the dependent variable. The selected sample set, independent variables, and dependent variables are input into the multiple linear regression module, and the least squares method is used for fitting. The constant term C, the regression coefficients A1 to A10 of the ten early warning indicators, and the t-statistics and p-values ​​of each coefficient are output to obtain the initial short-term forecast model, the expression of which is: ； Where Y is the dependent variable, i.e., the forecast amount of hail weather, and C is a constant. The regression coefficients of the ten early warning indicators are given. The independent variables are ten early warning indicators, namely, the Saud index, the K index, the lifting index, the storm intensity index, the convective effective potential energy, the 0°C layer height, the 0~-20°C layer thickness, the 0~6km deep vertical wind shear, the 500~700hPa temperature difference, and the 700hPa specific humidity. The regression coefficients of the ten early warning indicators are all in the range of [-1, 1].

5. A short-term, short-duration integrated hail weather warning method for the FAST area according to claim 4, characterized in that, The initial short-term forecast model was tested for significance using stepwise regression, resulting in an optimized short-term forecast model, including: Based on the above fitting results, the stepwise regression module in the statsmodels library is called. The significance is determined by the t-statistic and p-value of each independent variable in the initial short-term forecast model. The significance level for the introduced variables is set to p < 0.05, and the removal level is p > 0.

01. After iterative execution of the algorithm, the optimized model is automatically output. The optimized short-term forecast model is as follows: ； Where X1, X2, X3, and X4 represent the vertical wind shear of 0–6 km, the lifting index, the temperature difference of 500–700 hPa, and the storm intensity index, respectively, and A1, A2, A3, and A4 represent the regression coefficients of the vertical wind shear of 0–6 km, the lifting index, the temperature difference of 500–700 hPa, and the storm intensity index, respectively.

6. A short-term, short-duration integrated hail weather warning method for the FAST area according to claim 5, characterized in that, The short-term forecast model is validated based on historical hail weather radar data from conventional ground observation data and a selected sample set. The optimal hail weather forecast threshold is then output, including: The historical hail weather radar data after quality control and the selected sample set were substituted into the short-term forecast model; A formula for calculating the discrete true value of hail weather is constructed based on binary classification labels. The formula for calculating the discrete true value is as follows: ； These are the discrete true values ​​for hail weather. Compare the continuous forecast value Y calculated by the short-term forecast model with the discrete true value Y. i The accuracy rate of each preset forecast threshold to be determined is statistically analyzed to obtain the optimal forecast threshold for hail weather.

7. The short-term, short-duration integrated hail weather warning method for the FAST area according to claim 1, characterized in that, Quality control includes: Differential phase shift defolding: This involves checking the continuity of the radial differential propagation phase shift before and after the phase shift to ensure that the measurable range is limited to a 180° interval. Filtering: This involves performing comprehensive wavelet denoising on the defolded differential propagation phase shift and calculating the differential propagation phase shift rate using the least squares fitting method. Attenuation correction: An adaptive constraint algorithm is used to correct the attenuation of the horizontal reflectivity factor and the differential reflectivity factor, and adaptive Barnes interpolation is performed on various radar parameters.

8. A short-term, short-duration integrated hail weather warning method for the FAST area according to claim 1, characterized in that, The preset conditions for identifying strong convective cells include: combined reflectivity ≥ 45 dBZ; vertical cumulative liquid water content > 10 kg / m²; particle classification identifying high-density graupel or hail; and obvious mesocyclone characteristics, i.e., shear velocity ≥ 15 m / s.

Images